Trial design
Probiotic lactobacilli and bifidobacteria in antibiotic-associated diarrhoea and Clostridium difficile in the elderly was a prospective, multicentre, pragmatic, two-armed, double-blind, randomised, placebo-controlled trial with equal randomisation.
Approvals obtained
The Research Ethics Committee (REC) for Wales approved the study on 27 November 2008 (number 08/MRE09/18).
A clinical trial authorisation (CTA) was granted for the live bacterial preparation by the Medicines and Healthcare products Regulatory Agency (MHRA) on 6 October 2008 (CTA number AAD-CDD-001). The details of the REC for Wales, local RECs, competent authorities and research and development department approvals are provided in Appendix 1. The trial was assigned the International Standard Randomised Controlled Trial Number (ISRCTN) ISRCTN70017204 and EudraCT number 2007–002876–32.
Trial sites
Inpatients were recruited from medical and surgical wards of secondary care hospitals from two NHS regions: the Abertawe Bro Morgannwg University Health Board (ABMUHB) in South West Wales (Singleton, Morriston and Princess of Wales hospitals), and the County Durham and Darlington Foundation Trust (CDDFT; the University Hospital of North Durham and Darlington Memorial Hospital). Details of the trial sites are included in Appendix 2.
Participant inclusion criteria
People aged ≥ 65 years who were admitted to hospital and had been exposed to one or more antibiotics within the preceding 7 days, or were about to start antibiotic treatment, were eligible to join the study. Approval to invite the patient to participate in the study was secured from the patient’s consultant.
Participant exclusion criteria
People were excluded if they:
already had diarrhoea, which was defined as three or more watery or loose stools (Bristol Stool Form Scale types 5–7)
38 in the preceding 24 hours
were sufficiently immunocompromised to require isolation and/or barrier nursing
had a severe illness requiring high-dependency or intensive care
had a prosthetic heart valve
had suffered from CDD in the previous 3 months
had inflammatory bowel disease that had required specific treatment in the previous 12 months
had suspected acute pancreatitis, which was defined as abdominal pain with serum amylase or lipase greater than three times the institutional upper limit of normal
had a known abnormality or disease of mesenteric vessels or coeliac axis that may compromise gut blood supply
had a jejunal tube in situ or were receiving jejunal feeds
had a previous adverse reaction to probiotics
were unwilling to discontinue their existing use of probiotics.
Investigational medicinal products
The active preparation consisted of a vegetarian capsule containing 6 × 1010 live bacteria as lyophilised powder comprising two strains of L. acidophilus [CUL60, National Collection of Industrial, Food and Marine Bacteria (NCIMB) 30157 and CUL21, NCIMB 30156; 3 × 1010 colony-forming units (cfu)] and two strains of Bifidobacterium (B. bifidum CUL20, NCIMB 30153 and Bifidobacterium lactis CUL34, NCIMB 30172; 3 × 1010 cfu). Obsidian Research Ltd, Port Talbot, UK, prepared the investigational medicinal products (IMPs) according to the randomisation schedule. The organisms were available commercially through BioCare® Ltd (Lakeside, Birmingham, UK) and Pharmax (Bellevue, WA, USA). The placebo capsules were identical to the probiotic capsules but contained inert maltodextrin powder. The dose was one capsule per day with food and, where possible, between antibiotic doses, for 21 days.
Each participant was provided with a bottle labelled with his or her unique study number and containing 21 capsules. The IMPs were given to participants on the day of recruitment and administration during the hospital stay was supervised by the research nurses in ABMUHB and by the ward nurses in CDDFT. To prevent cross-contamination, strict hygiene procedures were followed, for example where capsules were opened, and the contents were mixed with fluids for administration to participants with difficulty swallowing. Although the live bacteria had a shelf life of 2 months at room temperature, participants were instructed to keep the capsules in the refrigerator following discharge from the hospital and encouraged to complete the 21-day course.
In vitro antibiotic testing demonstrated that the lactobacilli and bifidobacteria were sensitive to broad-spectrum antibiotics. All four strains were sensitive to ceftriaxone, chloramphenicol, erythromycin, linezolid and tetracycline.
After recruitment of the first 50 participants, a research nurse perceived a slight difference in the colour of the powder contained in the probiotic and placebo capsules, which were transparent. Therefore, the IMPs were resupplied in opaque capsules according to the original random allocation sequence. No unmasking of participant allocation occurred.
To ensure the quality of the IMPs, unused capsules were collected from participants on an opportunistic basis (e.g. from participants who withdrew before completing the 21-day course). A laboratory independent of the research team performed quantitative bacterial culture. The findings were sent to the independent statistician to check against the randomisation sequence and for the total count of viable organisms.
Objectives
Primary objectives: to determine the clinical effectiveness and cost-effectiveness of a live preparation of two strains of lactobacilli and two strains of bifidobacteria in the prevention of AAD and CDD in people aged ≥ 65 years who were exposed to oral or intravenous antibiotics and who were representative of patients admitted to secondary care NHS facilities in the UK.
Secondary objectives: to assess the effect of the probiotic on the duration and severity of AAD, the acceptability of the probiotic preparation, serious adverse events (SAEs) of the probiotic preparation and its effect on QoL.
Outcomes
Primary outcomes: the occurrence of AAD within 8 weeks and CDD within 12 weeks of recruitment.
Secondary outcomes: the severity and duration of AAD; the severity and duration of CDD and incidence of recurrence within the follow-up period; the number of days with abdominal pain, bloating, flatus and nausea; the incidence of pseudomembranous colitis, the need for colectomy; well-being and QoL; duration of hospital stay; frequency of SAEs; the acceptability of the live bacterial preparation (to identify if the participants had any difficulty taking the bacterial preparation vs. the placebo); the viability of the bacterial preparation at point of administration and death.
Sample size
Based on a review of previous clinical trials, we estimated that AAD would occur in 20% and CDD in 4% of participants in the placebo arm. The detection of a 50% reduction in CDD to a frequency of 2% in the active arm with 80% power at the 5% significance level required 2478 subjects (1239 in each arm). This number of participants would provide a power of > 99% to detect a 50% reduction in the risk of AAD (to 10% frequency) and a power of 90% to detect a 25% risk reduction in AAD (to 15% frequency) in the active arm at the 5% significance level. We planned to recruit 2974 participants to allow for 10% drop-outs and 10% loss to follow-up due to deaths unrelated to diarrhoea.
Randomisation
The independent statistician prepared a random allocation sequence using blocks of variable sizes and stratified by hospital using SAS® PROC PLAN, version 9.1 (SAS Institute Inc., Cary, NC, USA). The sequence allocated participants in a 1 : 1 ratio to the two arms of the study.
Blinding
The allocation sequence was not available to any member of the research team until databases had been completed and locked. In view of the established safety record of lactobacilli and bifidobacteria27 and the sensitivity of the organisms used in this study to broad-spectrum antibiotics, there was no provision for the emergency unblinding of participants. Therefore, copies of the random allocation sequence were not held at the recruitment sites.
Recruitment
The research nurses received training in ‘Good Clinical Practice’ and were fully conversant with all aspects of the trial. Specific training was given in assessment of patient eligibility, recruitment and obtaining informed consent, the trial protocol, reporting of adverse events and collecting information. Presentations were made to clinical, nursing and pharmacy staff to ensure their familiarity with the purpose and conduct of the trial. Permission to approach patients and invite them to join the study was obtained from hospital consultants.
Research nurses visited wards daily to identify eligible patients and provide them with verbal and written information (see Appendix 3). Patients were revisited either later the same day or the next day after they had had the opportunity to discuss with relatives/carers and health-care personnel whether or not they wished to join the study. For people unable to provide consent, information was provided to their relatives/carers and assent sought (see Appendix 3). Although reasons for not joining the study were requested, patients and their relatives/carers were free to decline to participate without giving a reason. Patients in whom consent or assent was obtained were allocated the next unique study number in the random sequence for that site and the research nurses provided them with the corresponding trial preparation.
Baseline assessment
Information recorded at recruitment included basic demographic characteristics, use of cigarettes and alcohol, source of admission, principal diagnosis or reason for admission, comorbid illnesses, duration of hospital stay prior to recruitment, non-antibiotic drug treatment, indication for antibiotic treatment and antibiotics prescribed (see Appendix 4).
Participant follow-up
Participants were visited daily by the research staff during hospital stay. Changes to antibiotic treatment, gastrointestinal symptoms (including the presence of diarrhoea), compliance, difficulties taking the IMPs and occurrence of SAEs were recorded onto standard forms (see Appendix 4). The same information was requested weekly, after discharge from hospital, via a telephone questionnaire and continued for 8 weeks post recruitment. In addition, participants were encouraged to contact a named member of the research staff to report potential SAEs at any time during follow-up. Review of laboratory data regarding stool assays was continued until 12 weeks after recruitment.
Trial completion
Participants had completed the trial when they:
A chart showing participant flow through the study is included as Appendix 5.
Measurement of primary outcomes
Diarrhoea was defined as three or more loose stools (stool consistency 5–7 on the Bristol Stool Form Scale)38 in a 24-hour period. Diarrhoea was also diagnosed in participants with frequent stools that they described as loose but who were unable to describe stool consistency using the Bristol Stool Form Scale. The presence of diarrhoea according to these criteria was confirmed by the research nurses during admission with either the patient, their carers or a member of the medical staff. After discharge, this was confirmed during telephone interview.
Stool samples for analysis were collected only during episodes of diarrhoea, including diarrhoea that occurred after discharge from hospital. Stools were analysed for diarrhoeal pathogens according to routine NHS laboratory practice. Analyses included bacterial culture for Salmonella spp., Shigella spp., Campylobacter and Escherichia coli 0157 and wet film for ova, cysts and parasites. Detection of viruses depended on the clinical context (e.g. suspected norovirus outbreak). In ABMUHB, C. difficile toxins were detected by an in-house tissue culture assay with confirmation by enzyme immunoassay Premier™ Toxins A&B (Meridian Bioscience, Inc., Cincinnati, OH, USA). In CDDFT, toxins were detected by the VIDAS®
C. difficile A&B (bioMérieux SA, Marcy l’Etoile, France). To improve the detection of CDD from June 2010, CDDFT also introduced detection of the C. difficile antigen glutamate dehydrogenase C. DIFF QUIK CHEK® (TECHLAB® Inc., Blacksburg, VA, USA) to be used in conjunction with the toxin assay. Further stools samples were requested if a cause of diarrhoea was not identified and especially if there was clinical suspicion of CDD. Stools positive for C. difficile toxin were cultured and the isolates sent to a central reference laboratory for ribotyping.
Antibiotic-associated diarrhoea was defined as occurring in association with antibiotic therapy but without diarrhoeal pathogens detected on routine laboratory analysis or an alternative explanation (e.g. laxative treatment). Among patients with AAD, CDD was diagnosed in those with stools testing positive for C. difficile toxins.
Measurement of secondary outcomes
Information regarding the severity and duration of AAD and CDD, number of stools per day and stool consistency, incidence of recurrence of CDD within the follow-up period, the number of days with abdominal pain, bloating, flatus and nausea, duration of hospital stay and acceptability of the live bacterial preparation was collected by the research nurses during follow-up (see Appendix 4). CDD was investigated and managed according to the current hospital practice and clinical and laboratory information from clinical records was recorded (see Appendix 4). Information regarding the occurrence of pseudomembranous colitis, the need for colectomy and death was also collected from the patient’s case records. The information from follow-up and patient case records was used to classify the severity of CDD according to UK national guidelines (see Appendix 6).39 The severity of CDD was classified independently by two assessors, WH and SA, and differences resolved by discussion.
Quality of life was assessed by the generic Short Form questionnaire-12 items version 2 (SF-12 v2) administered by research nurses at baseline and at 4 and 8 weeks. SF-12 v2 QoL subscales (social function, role function–emotional, role function–physical, general health, mental health, bodily pain, vitality, physical functioning), physical component summary (PCS) and mental component summary (MCS) scores were calculated and quality assured according to the QualityMetric Incorporated guidance40 with imputation of missing scores using the SF-12 v2 Missing Data Estimator software where possible.40 SF-12 v2 subdomain, PCS and MCS scores were allocated a value of 0 for the lowest/worse score and 100 for the highest/best score.
Serious adverse events were identified and reported according to standard guidelines.41 Suspected unexpected serious adverse reactions were to be reported immediately to the independent safety monitor and, if appropriate, the ethics committee in accordance with local and national requirements, which were identified as:
any manifestation of infection (e.g. abscess, bacterial endocarditis, bacteraemia) in which lactobacilli or bifidobacteria were isolated from pathological specimens
the development of pancreatitis, which was defined as abdominal pain with serum amylase or lipase concentration equal to or greater than three times the institutional upper limit of normal
the development of multiorgan failure
the development of bowel ischaemia.
Additional data collected
Information was collected to identify subgroups of participants who may be at increased risk of AAD and CDD. This included potential risk factors at baseline, antibiotic treatment and duration of hospital stay.
Data management
All data were collected on standardised forms that were checked for missing values by the trial manager. Routine laboratory records were accessed to record results of stool analyses. Data were entered into Microsoft Access® (Microsoft Corporation, Redmond, WA, USA) and included range checks and double entry. Databases were compared using Microsoft Excel® (Microsoft Corporation, Redmond, WA, USA) to identify data entry errors.
Antibiotics were classified according to British National Formulary categories (see Appendix 7).42 Indications for antibiotic treatment were classified according to the system organ class (SOC) terminology of the Medical Dictionary for Regulatory Activities (MedDRA).43 Participants were allocated to each SOC for either suspected or proven infection of that organ or system. Antibiotic treatment for suspected infection where no system or organ was identified was classified as ‘suspected sepsis but site unclear’. SAEs were coded according to the most appropriate preferred terms (PTs) of the MedDRA.43
After data cleaning, databases were locked and forwarded to the trial statistician for analysis. Initial analysis according to the randomisation sequence identified the two arms of the study as only ‘A’ or ‘B’. The identity of the two arms was only revealed after the Data Monitoring and Ethics Committee had reviewed these data and approved the analysis.
Statistical analysis
Demographic and baseline data were summarised by recruitment hospital and treatment group. Continuous variables were summarised using number of observations, median and interquartile range (IQR) and categorical variables by the number and percentage of events.
Analysis of primary and secondary end points was performed in a modified intention-to-treat (ITT) population that excluded a small number of subjects who withdrew shortly after randomisation and did not have follow-up data. The chi-squared test or Fisher’s exact test was used to compare proportions. Risk difference and RR together with the 95% CIs were calculated using a generalised linear model that included treatment as a single predictor. Similarly, CIs for the ORs of AAD and CDD were estimated from logistic regression models. Secondary continuous outcomes with no repeated measurements were summarised using number of observations, median and IQR. The t-test or Mann–Whitney method were used to compare continuous variables. Duration data were summarised by median and IQR and compared using the Mann–Whitney method. No transformation was used for any variables.
Analysis of primary end points was also performed by logistic regression model adjusting for the following prespecified baseline characteristics and potential risk factors for AAD that may be likely to affect the occurrence of the primary end point:
centre
age
sex
antibiotic class
duration of antibiotic treatment
antacid therapy (including PPI treatment)
NGT in situ
previous gastrointestinal surgery
recent previous hospital admission
duration of hospital stay.
We intended to include all 10 prespecified variables in the logistic model but some were not entered because their effects were inestimable. A per-protocol (PP) population excluded participants who did not receive any IMP doses or in whom compliance was unclear. Additional analyses also assessed probiotic effect on primary outcomes in participants who took all 21 doses, 14 or more doses, or seven or more doses of the IMP. Analysis methods for the PP population were similar to those described for the modified ITT population. All analyses were performed on both the modified ITT and PP populations.
SAS version 9.2 was used for data analyses.
Quality of life analysis
The main analysis compared the change from baseline in SF-12 v2 PCS and MCS composite scores at 4 weeks in the two study arms. We also compared SF-12 v2 PCS and MCS composite scores at 8 weeks and scores for individual SF-12 v2 subdomains (social function, role function–emotional, role function–physical, general health, mental health, bodily pain, vitality and physical functioning) at 4 and 8 weeks. Composite scores were compared using mixed model analysis using the SF-12 v2 score at baseline as a covariate, treatment, visit, interaction between the treatment and visit as fixed effects, and subject as a random effect. During the trial, some subjects dropped out, resulting in some incomplete observations. These incomplete observations were not computed but were assumed to be missing at random in the mixed model analysis. The treatment difference, together with the 95% CI at each visit, was derived from the mixed model.
Economic analysis
The cost-effectiveness evaluation was undertaken from the perspective of the UK NHS. Costs were assigned to the resources utilised by each participant. These consisted of the bacterial preparation, staff time involved in administering the preparation, treatment relating to potential adverse events, the assessment of cases of diarrhoea (stool collection and culture/toxin assay, endoscopy) and diarrhoea management costs (laundry, antibiotics, increased hospital stay and comorbidities). Unit costs (cost year 2011) were derived from published information44–46 and through discussions with relevant clinical and finance department staff. Missing data were addressed using the imputation-based method for QoL data and censored data relating to costs using the weighted cost method with known cost histories.47 In view of the short timescale of the project, there was no discounting of the costs or benefits. Costs and benefits would have been discounted at the conventional rate of 3.5% if the time scale of the follow-up had exceeded 1 year.
Cost differences between the two arms of the study were determined and used in conjunction with differences in outcomes in undertaking a cost–consequences analysis, with cost per case of AAD averted as the primary outcome measure. We planned to undertake subgroup analyses to determine the relative cost-effectiveness of preventative strategies in different risk groups as indicated in the covariate analysis. In addition, a cost–utility analysis considered the differences in costs between the two study arms and differences in quality-adjusted life-years (QALYs) derived from European Quality of Life-5 Dimensions (EQ-5D) responses.
Resource use and costs
Health-care contacts
The number of health-care contacts, duration of initial hospital stay, days spent in care facilities and number and duration of readmissions were recorded routinely within the trial. If the discharge date was not known, the end of follow-up was assumed to be the discharge date. A weighted unit cost of £334.17 was applied for every day a participant spent in hospital (NHS reference costs 2011; extrapolated).44 For the base-case analysis, all other health-care contacts were treated as general practitioner (GP) visits and published care home costs were applied on a daily rate basis.45
Antibiotics
Antibiotic use was collected throughout the 8-week follow-up period and costed using published sources.46 Staff time was assumed to be 5 minutes for the administration of oral antibiotics and 20 minutes for intravenous or intramuscular antibiotics. As 66% of doses during the study period were delivered orally and 34% intravenously, staff cost per dose was weighted to 10.1 minutes per dose and costed at £16.33 per dose.45 Missing data on antibiotic dose were replaced by the most common value. If no data were available on number of doses per day or duration of antibiotic course, it was assumed that the patient was receiving a full course of the recorded antibiotic.
Intervention implementation
For the cost-effectiveness analysis, it was assumed that every patient would receive a course of 21 oral capsules containing the probiotic formulation at a retail cost of £10 and that capsules that were not taken by the participant would go to waste (as a high proportion of participants finished their course at home after hospital discharge). While the patient was in hospital, staff time of 5 minutes was allocated for administration of each dose. The number of days in hospital (and thus number of nurse contacts) was calculated individually for each patient according to his or her intervention start and hospital discharge dates. Nursing time was allocated even if the patient declined the intervention, as time would have been used for patient interaction and assessment of the situation. No staff time was allocated after the patient withdrew or died. Nursing time was costed at £8.08 per 5 minutes.45 The cost of placebo formulation and staff cost for administration in the control group were excluded from the cost-effectiveness analysis as ‘routine use’ was considered.
Episodes of diarrhoea
Costs regarding health-care resource use, antibiotics and increase in length of hospital stay associated with diarrhoea were collected routinely during the trial. Other costs, such as diagnostics, clinical review, cleaning, laundry and isolation, were sourced from outside the trial setting. Resource use and costs of laboratory tests for C. difficile detection were obtained using a microcosting approach based on internal standard operating procedures and discussions with key laboratory staff and purchasing officers. Costs of other diagnostic tests (Salmonella, Shigella, Campylobacter, E. coli 0157 and norovirus) were taken from the literature.48 Published reference costs44 were used to estimate the costs of diagnostic (endoscopy, abdominal computerised tomography and radiography) and therapeutic (colectomy) procedures, which were then weighted to reflect the probability of the event occurring in a population suffering from diarrhoea based on recent publications.49,50 Costs of patient assessment, including for review of antibiotic treatment and nutritional requirements, and infection control measures were based on discussions with infection control nurses, clinicians and microbiologists. The medical team in the base case was assumed to consist of the treating clinician (costed as registrar), a gastroenterologist, a microbiologist, an infection control nurse, a ward nurse and a pharmacist. Staff time was estimated to be 45 minutes for the registrar and 15 minutes for the other professionals and unit costs were obtained from the literature.45 Cleaning after patient discharge from the hospital or relocation after a diarrhoea episode was costed based on discussions with key members of infection control staff and includes domestic staff time, specialist cleaning agents (hypochlorite; TUFFIE 5®, Vernacare UK, Bolton, UK) and special cleaning equipment as well as laundry, steam cleaning and use of an autoclave. Costs were obtained from hospital human resource and purchasing departments, wholesalers, published literature51 and microcosting. All costs were allocated once per patient diarrhoea episode.
Daily costs of diarrhoeal disease included daily cleaning and bed and ward closures. Costing of daily cleaning included domestic staff time, specialist cleaning agents and special cleaning equipment. Costs were obtained from human resources and purchasing departments as well as wholesalers. A lost bed-day due to closures was costed at £334.17 by published reference costs44 and weighted for specialties (e.g. whether or not the bed is in intensive care) and activity.44 Data on frequency of ward closures due to CDD and number of positive cases identified per year were obtained from discussions with key staff and Public Health Wales reports.52 It was assumed that 1 in 115 cases resulted in an outbreak and subsequent ward closure. Based on an average ward size of 28 beds in Singleton and Morriston hospitals, ward closures could cost up to £9356.76 per day and occur in 0.87% of positive cases. Thus, a weighted cost of £81.40 was applied to each diarrhoea case per day to account for potential ward closures.
The cost per stool included disposables such as gloves and aprons, laundry and staff time for patient check-ups, spot cleaning and changing of beds. Costing of spot cleaning included nursing time, specialist cleaning agents and special cleaning equipment. Data on number of stools per day, duration of diarrhoea episode in days and whether or not the episode was managed in hospital were collected routinely as part of the trial. These data were used to calculate the cost of a diarrhoea episode individually for each patient by applying the one-off episode costs (microbiology, review, procedures, end cleaning) and adding daily and per stool costs according to duration and stool frequency. No cost was applied to participants whose diarrhoea was managed entirely at home. Episodes managed in care homes were treated as hospital episodes.
Cost-effectiveness analyses
Patient responses from EQ-5D questionnaires at baseline and after 4 and 8 weeks were translated into utility scores using a scoring algorithm. We planned to use health and cost outcomes to calculate the cost of probiotics per QALY gained in a cost–utility analysis, to obtain the cost per case of diarrhoea averted in a cost-effectiveness analysis and to present and compare outcomes in tabular form in a cost–consequences analysis. The results of cost-effectiveness analyses were expressed as incremental cost-effectiveness ratios (ICERs) and calculated by dividing the cost difference between the two alternatives being compared by the difference in the effect/benefit. In cost–utility analysis, the effect was expressed in QALYs. The cost per QALY was calculated by dividing the difference in costs by the difference in QALYs for each comparison.
Generally, the National Institute for Health and Care Excellence considers an intervention cost-effective if one of the following applies.53
The intervention is less costly and more clinically effective than all other relevant alternatives. In this case, no ICER is calculated as the strategy in question dominates the alternatives.
The intervention has an ICER of < £20,000 per QALY compared with the next best alternative. This means that an investment of up to £20,000 in order to achieve an additional QALY is considered cost-effective.
Sensitivity analysis
Sensitivity analyses investigated the robustness of the results to changes in estimated costs and outcomes and probabilistic sensitivity analysis used bootstrap resampling to determine the probability that preventative strategies were within certain thresholds. We planned to assess the budgetary impact from a NHS perspective of adopting a policy of administering the bacterial preparation to prevent or ameliorate AAD and CDD in the target population.
During univariate sensitivity analysis, all ICERs were recalculated after changing the value of a range of parameters individually to estimate the robustness of the ICER (). Prolonged inpatient stay is the main cost driver when considering the cost of diarrhoea.13,15,16 Other potentially important cost differences between the probiotic and placebo arms included staff time, which is naturally subject to variation, and diarrhoea-associated costs such as cleaning, laundry, microbiology, assessment, diagnostic and therapeutic procedures. As these costs were thought to make up a large proportion of the total health-care costs, the impact of changes to parameters contained within these costs was evaluated.
Parameter changes for univariate sensitivity analysis
Probabilistic sensitivity analysis with changes to the values of all chosen parameters (usually within the 95% CI or a reasonable, defined range) calculated the probability that the intervention was cost-effective when all uncertainty associated with the individual parameters was considered. Results of the probabilistic sensitivity analysis were expressed as per cent probability that the intervention was cost-effective. Cost-effectiveness acceptability curves (CEACs) were generated to depict the probability of the intervention being cost-effective at different willingness-to-pay thresholds.
